AI Token expenditure index drops nearly 20%, with signs of cooling in users' marginal willingness to pay

```

The global AI trading market is facing a cooldown in a key observation indicator. The core index that measures spending on AI Token usage has declined after a rapid surge, prompting the market to reassess whether this AI cycle, driven by massive capital expenditure, remains solid—both demand strength and pricing power are now under stricter scrutiny.

According to Bloomberg, Silicon Data's LLM Token Expenditure Index, launched last December, nearly doubled at one point, but has fallen by nearly 20% since peaking in May this year. This index is seen as the most direct market indicator reflecting willingness to pay for AI services and marginal demand changes, and its weakening is causing a reassessment of the pace of AI commercialization.

Some investors believe this change may indicate that AI companies are entering a more cost-sensitive competitive environment, and pricing power may be eroding. Veteran investor Louis Navellier pointed out that signs are showing some users are starting to limit usage due to cost pressures. Meanwhile, market rumors about OpenAI possibly delaying its IPO are also seen as indirect evidence of lingering profitability challenges.

However, the decline in the index does not mean AI services themselves are "getting cheaper." Silicon Data emphasizes that this indicator is the result of the interaction between price and usage, closer to a proxy for "marginal willingness to pay". Therefore, the driving factors behind its trend are different, and the corresponding market implications may also be entirely different.

Rising Divergence: Cooling Demand or Pricing Structure Shift?

There are clear market divergences in how the decline in the index is interpreted.

An optimistic interpretation suggests that since 2023 token prices have dropped by more than 90%, significantly lowering the usage threshold and thus expanding overall expenditure. In this framework, the index's temporary decline reflects structural adjustment in demand rather than overall weakening, and the logic of AI expansion still stands.

But the pessimistic view suggests that this may mean users’ willingness to pay has reached a temporary limit. Allianz Research points out that there is currently about a 46% growth gap between AI investment and actual sales, higher than the 32% deviation during the 2001 telecom bubble. Against this backdrop, any weakness on the demand side could amplify valuation pressure.

Louis Navellier also mentioned that enterprises are facing cost constraints in their use of AI services, starting to restrict "unlimited usage," which is seen by some market participants as an early signal of declining demand elasticity.

Capital Expenditure Logic Remains, but Structure Is Changing

Despite fluctuating demand signals, the AI infrastructure investment cycle has not reversed significantly. Market data shows that high-end GPUs and high-bandwidth memory continue to be in short supply, expected to last until 2026, and possibly extend to 2028.

However, market attention is shifting from training to inference. Reports note this means the structure of computing power demand is migrating; the proportion of high-end training GPUs may decline, while demand for inference-optimized hardware is relatively increasing, thus changing the industry value chain's beneficiary structure.

This change does not directly constitute a bearish judgment on the chip industry, but may indicate that growth drivers are shifting from "expansion of high-end computing power" to "structural redistribution of demand."

Policy and Regulatory Variables Increase Pricing Complexity for Companies

Beyond demand itself, the regulatory environment is also impacting the commercialization path of AI products.

Recently, US regulators have requested adjustments to the release pace of certain models and relaxed access restrictions for some Anthropic PBC models. Meanwhile, the EU AI Act is imposing mandatory evaluations and stricter transparency requirements on advanced models.

These policy changes do not directly restrict pricing but add pressure from deployment and compliance costs, making companies more likely to prioritize cost optimization when distributing workloads across different models. This trend may indirectly affect the pricing power and usage share of high-end models.

Risk Warning and DisclaimerThe market contains risks, and investment requires caution. This article does not constitute personal investment advice and does not consider the unique investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions contained herein are suitable for their specific circumstances. Investment based on this is at your own risk. ```